Missing Data in Afghanistan: A Latent Class Examination of Item Nonresponse

• Activity Number: 138
• Type: Contributed
• Date/Time: Monday, August 4, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Survey Research Methods Section
• Author(s): Ann Arthur*+ and Jill Heese and Allan McCutcheon
• Companies: Gallup Research Center and Gallup Research Center and University of Nebraska-Lincoln
• Keywords: Afghanistan ; latent class analysis ; non-response ; missing data

Abstract

With its first presidential election in 2004, Afghanistan is still in the infancy of its democracy. Although a key component of building a strong democracy is citizen participation (Aristotle, n.d./1995), Afghan women give fewer opinions to political questions than men (e.g., "Don't know") (Heese & Arthur, 2012). Despite gender-matched interviews, this behavior is not uncommon in Afghan culture. This research examines the frequency with which Afghan women give substantive survey responses using latent class analysis (McCutcheon, 1987). Using a probability-based sample of 1,000 Afghan women from the 2012 Gallup World Poll, four political questions are analyzed. Preliminary analyses indicate three classes: women with the highest probabilities of answering all questions (70.8%), those with moderate probabilities of answering questions (16.1%), and a third class with a mix of high and low probabilities (13.2%). This research then describes membership in the classes with key demographic variables. The results identify challenges related to data collection in this region and how those challenges undermine efforts to obtain good estimates on important issues.